AFAIK blocks are defined by their offset and length alone, correct? You can get those details for a given source via DFSClient.getBlockLocations() and then perhaps construct a manual FileSplit object with the length details to read out a single block (Am not sure if this would handle records properly).
2010/12/16 Petrucci Andreas <[email protected]>: > > hi there, i want to aski if hdfs api supports reading just a specific block > of a file (of course if file exceeds the default block size). for example is > it possible to read/fetch just the first of the third block of a specific > file in hdfs? does the api supports that? > -- Harsh J www.harshj.com
